Performance Comparison of PhaseNet and Blockly Earthquake Transformer in Automatic First Arrival Picking on the Cianjur Earthquake

On November 21. 2022. an earthquake with a magnitude of 5.6 struck Cianjur West Java. causing extensive damage. By giving the public early warnings. an Earthquake Early Warning System (EEWS) can assist lessen the impact of earthquakes. The seismic phase selecting system. which works to detect and identify seismic waves originating from earthquakes. is one crucial part of EEWS. In this work. the effectiveness of PhaseNet and Blockly Earthquake Transformer (BET) . two seismic phase selecting techniques. is compared using Cianjur earthquake data. PhaseNet is a deep learning model that recognizes and classifies seismic waves using a convolutional neural network (CNN). A deep learning model called Blockly Earthquake Transformer employs transformer architecture to recognize and identify seismic waves. The average computational speed of PhaseNet is 8.25 seconds. while that of BET is 6.13 seconds. The results indicate that PhaseNet generates average P_probability values of 0.56 and average S probabilities of 0.53. which are higher than those of the BET. which generates average P_probability values of 0.40 and average S probabilities of 0.32. It is evident from the Cianjur earthquake data that both approaches perform well in terms of recognizing and detecting seismic waves. Compared to the BET. PhaseNet is marginally more accurate. In terms of computing time. the BET is more effective and faster.